Cytochrome P450 (P450) enzymes play an important role in the metabolism of exogenous and endogenous molecules. CYP3A4 is a major form of P450 expressed in adult liver and small intestine, where it participates in the biotransformation of numerous clinically important drugs. A structurally related isoform in the CYP3A subfamily, CYP3A5, is polymorphically expressed in the liver, small intestine, and kidney (Wrighton et al., 1990; Haehner et al., 1996; Paine et al., 1997). Recent investigations have shown that hepatic CYP3A5 is present at readily detectable levels in 10 to 30% of whites and approximately 50% of African Americans (Hustert et al., 2001; Kuehl et al., 2001; Lin et al., 2002). Moreover, it seems that the relative abundance of CYP3A5 in some human livers can approach that of CYP3A4 (Lin et al., 2002), although some investigators have reported uniformly lower relative CYP3A5 expression in their tissue bank (Westlind-Johnsson et al., 2003).

Because CYP3A5 exhibits an overlapping substrate specificity with that of CYP3A4, it may contribute significantly to the metabolic clearance of CYP3A substrates in people carrying the wild-type CYP3A5^*1 allele, although in vivo data as well as in vitro evidence are conflicting. For example, Floyd et al. (2003) found no effect of the inactivating CYP3A5^*3 and CYP3A5^*6 alleles on midazolam (MDZ) clearance in a mixed population of healthy white and African American adults. Similarly, negative findings were recently reported by Yu et al. (2004). In contrast, Wong et al. (2004) reported a significantly higher midazolam clearance for a population of largely white cancer patients carrying the wild-type CYP3A5^*1 allele compared with patients with a homozygous CYP3A5^*3 genotype. Interestingly, both Floyd et al. (2003) and Yu et al. (2004) observed a significant effect of the CYP3A5 polymorphism on the magnitude of a drug-midazolam interaction in a direction consistent with reduced inducibility or inhibition of CYP3A5, respectively. In the case of the immunosuppressive agent tacrolimus, a drug also metabolized by both CYP3A4 and CYP3A5 (Bader et al., 2000), the data are much more consistent. Multiple independent groups of investigators have shown that patients carrying the CYP3A5^*1 allele exhibit a lower dose-adjusted steady-state blood level (i.e., higher oral clearance) than do patients with a homozygous CYP3A5^*3 genotype (Hesselink et al., 2003; Thervet et al., 2003; Haufroid et al., 2004). This suggests that the in vivo significance of the CYP3A5 polymorphism may be substrate-dependent.

Although CYP3A5 metabolizes most CYP3A4 substrates, it often displays differences in product regioselectivity and lower metabolic activity compared with CYP3A4. For example, several investigators report reduced CYP3A5 metabolic capacity compared with CYP3A4 toward testosterone (TST) (Wrighton et al., 1990; Waxman et al., 1991), progesterone (Aoyama et al., 1989; Waxman et al., 1991), nifedipine (Wrighton et al., 1990), and erythromycin (ERM) (Wrighton et al., 1990). In contrast, other studies suggest that CYP3A5 catalyzes the metabolism of several CYP3A4 substrates with approximately equal or higher efficiency, including testosterone (Gillam et al., 1995), nifedipine (Aoyama et al., 1989; Gillam et al., 1995), erythromycin (Gillam et al., 1995), midazolam (Gorski et al., 1994), diltiazem (Yamaori et al., 2004), and lidocaine (LDC) (Bargetzi et al., 1989). More recently, Williams et al. (2002) reported an equal (midazolam 1′-hydroxylation) or reduced metabolic capability (oxidation of testosterone, alprazolam, triazolam, diltiazem, clarithromycin, estradiol, and nifedipine) for CYP3A5 compared with CYP3A4, and yet another group found that CYP3A5 has significantly lower metabolic activity than CYP3A4 toward midazolam, triazolam, testosterone, and nifedipine (Patki et al., 2003).

The apparent discordance in CYP3A experimental results obtained for the same substrate can be attributed in part to a variable enzyme source: membrane fractions from transfected bacteria (Yamazaki et al., 1999), yeast (Imaoka et al., 1996), human lymphoblastoid cells (Yamazaki et al., 1999), and insect cells (Yamazaki et al., 1999). In addition, the matrix employed can be different with respect to specific lipid mixtures, cholate, detergent, buffer, salt compositions, and components of the NADPH-regenerating system used (Yamazaki et al., 1996, 1997; Maenpaa et al., 1998). Even with the same P450 expression system, the contents of NADPH-dependent cytochrome P450 reductase (OR) and cytochrome b₅, which contribute to different CYP3A4 and CYP3A5 metabolic activities (Yamazaki et al., 2002; Yamaori et al., 2004), may vary.

Given this background, we re-evaluated the contribution of CYP3A5 to the hepatic metabolism of eight structurally diverse molecules. To accomplish this, we used commercially available baculovirus-infected insect cells expressing CYP3A4 or CYP3A5 with a comparable level of reductase activity, optimizing the metabolic rates by adding an equivalent amount of b₅ to both enzyme systems. In addition, we used human liver microsomes (HLMs) that were either deficient or high in CYP3A5 content to investigate the contribution of CYP3A5 in a more functionally relevant catalytic system.

Materials and Methods

Materials. Supersomes prepared from baculovirus-infected insect cells expressing human CYP3A4 and CYP3A5 [CYP3A4 coexpressed with reductase (with or without b₅) and CYP3A5 coexpressed with reductase] were purchased from BD Gentest (Woburn, MA). Human cytochrome b₅ and human NADPH-P450 reductase were purchased from PanVera Corp. (Madison, WI). Testosterone, flunitrazepam (FLZ), carbamazepine (CBZ), terfenadine (TFN), dextromethorphan (DXM), lidocaine hydrochloride, erythromycin, itraconazole (ITZ), 2-methyl carbamazepine, and carbamazepine 10,11-epoxide were obtained from Sigma-Aldrich (St. Louis, MO). 6β-Hydroxytestosterone and 11β-progesterone were obtained from Steraloids (Newport, RI). 3-Hydroxy-flunitrazepam and desmethylfunitrazepam were a gift from Hoffman-La Roche (Nutley, NJ). Ethoxyl-oxo-terfenadine was obtained from Aventis (Strasbourg, France). N-Methyl-N-(t-butyl-dimethylsilyl) trifluoroacetamide was obtained from Pierce Chemical (Rockford, IL). Hydroxy-itraconazole (OH-ITZ) was purchased from Research Diagnostics Inc. (Flanders, NJ). Terfenadine alcohol and carboxyl acid were gifts from BD Gentest. Midazolam, 1′-hydroxymidazolam, and ¹⁵N₃-midazolam were gifts from Hoffman-La Roche. Monoethylglycinexylidine (MEGX) was a gift from Abbott Laboratories (Evanston, IL). 3-Methoxymorphinan (3-MM) was kindly provided by Dr. Stephen Hall (Indiana University School of Medicine, Indianapolis, IN). [N-Methyl-¹⁴C]erythromycin was kindly provided by Dr. Paul Watkins (University of North Carolina, Chapel Hill, NC). All other chemicals and solvents used were of analytical grade and without further purification.

Microsomal Preparations. Human liver samples were obtained from an existing bank maintained by the University of Washington School of Pharmacy. The CYP3A expression and activity of these tissues have been thoroughly described (Lin et al., 2002). Microsomes were prepared as described previously (Paine et al., 1997). Protein concentrations were measured by the Lowry method (Lowry et al., 1951). Human liver microsomes were analyzed for immunodetectable CYP3A4 and CYP3A5 content as described previously (Lin et al., 2002). Two sets of microsomal preparations were chosen; group 1 expressed CYP3A4 predominantly with only trace or no detectable CYP3A5 (n = 10), and group 2 expressed both CYP3A4 and CYP3A5 (n = 10). Each liver in group 1 was matched to a liver in group 2 with respect to the CYP3A4 expression level.

CYP3A Preparations and Incubations. All incubations were conducted in duplicate, with the appropriate controls. Each incubation mixture (0.2-1.0 ml) contained 100 mM KPi (pH 7.4) and 1 mM EDTA. The enzymes were preincubated in a 37°C shaking water bath for 5 min prior to NADPH addition (final concentration, 1 mM) to initiate the reaction. The volume of organic solvent used to introduce substrate did not exceed 1% (v/v) of the total incubation volume. The formation of metabolites was linear with respect to incubation time and microsomal protein concentration. Total substrate consumption in each reaction was less than 15% of the initial substrate added.

In a pilot experiment, testosterone 6β-hydroxylation was measured to compare the catalytic activity of CYP3A4 coexpressed with reductase and b₅ (CYP3A4/reductase/b₅ molar ratio = 1:6.6:7.1) and CYP3A4 coexpressed with reductase (CYP3A4/reductase molar ratio = 1:0.31) supplemented with additional human reductase and b₅ to adjust the molarity to a comparable 1:6.6:7.1 ratio.

The metabolic activities of CYP3A4 and CYP3A5 toward the eight probe substrates were compared using Supersomes of CYP3A4 and CYP3A5 coexpressed with comparable reductase activity. For this experiment, both CYP3A4 and CYP3A5 preparations had approximately equal P450 content/reductase activity ratios and were supplemented with b₅ (CYP3A/b₅ molar ratio = 1:3). Table 1 summarizes the incubation conditions and analytical methods employed for each probe substrate.

Itraconazole Metabolism by P450 Isoforms. We recently reported that itraconazole is metabolized selectively by CYP3A4 and not by CYP3A5 (Isoherranen et al., 2004). Thus, although no kinetic comparison was performed for this molecule in the present study, we used the maximal itraconazole hydroxylation activity as an additional control (in addition to CYP3A4 protein content) for interliver differences in CYP3A4 activity. The maximal rate of itraconazole hydroxylation by liver microsomes was measured as described above. Briefly, each incubation was carried out in 200 μl of KPi with 0.05 mg/ml microsomal protein (0.01 mg in 0.2 ml of final volume). The final nominal ITZ concentration was 500 nM, which is ∼10-fold higher than the apparent K_m for the reaction. The reactions were initiated with the addition of NADPH (1 mM). After 2 min, the reactions were quenched with the addition of 200 μl of acetonitrile, and the formation of OH-ITZ was measured by liquid chromatography/mass spectometry (Table 1).

Other Assays. For human liver microsomes, NADPH-cytochrome c reduction activities were determined as described previously (Yasukochi and Masters, 1976). P450 reductase concentrations were calculated assuming a specific activity of 3.0 μmol of cytochrome c reduced per minute per nanomole of reductase, based on published data for purified human and rabbit reductase preparations (Parikh et al., 1997). The concentration of cytochrome b₅ was calculated spectrally following published methods (Omura and Sato, 1964a,b).

Data Analysis. Biotransformation of the eight probe substrates by heterologously expressed CYP3A4 and CYP3A5 was described by one of the following models: classic hyperbolic (eq. 1; Segel, 1975); sigmoidal (eq. 2; Segel, 1975); substrate inhibition (eq. 3; Kronbach et al., 1989); or biphasic saturation (eq. 4; Korzekwa et al., 1998). The kinetic parameters were estimated using nonlinear regression (WinNonlin; Pharsight, Mountain View, CA). The reported models were chosen based on results of the F test for nested models and the distribution of residuals.

For hyperbolic and substrate inhibition models, the unbound intrinsic metabolic clearance (CL′_int) was calculated as V_max/K_m. For the biphasic saturation model, CL′_int was estimated as V_max1/K_m1 (high-affinity, low-capacity site). For the sigmoidal model, we calculated the maximal intrinsic clearance (CL_max) by using eq. 5 (Segel, 1975; Houston and Kenworthy, 2000; Houston and Galetin, 2003).

Plots were generated using KaleidaGraph software (Abelbeck/Synergy, Reading, PA); t tests and Pearson correlation analysis of liver microsomal samples were performed using SPSS software (SPSS Inc., Chicago, IL). All of the comparisons employed a two-tailed unpaired t test.

Results

Kinetic Comparison of CYP3A4 Coexpressed with and without Cytochrome b₅. For 6β-OH TST formation, CYP3A4 coexpressed with OR and b₅ (CYP3A4 + OR + b₅) exhibited homotropic activation and was best fit by the sigmoidal model with a Hill coefficient of 1.45. When CYP3A4 coexpressed with reductase (CYP3A4 + OR) was supplemented with additional reductase and b₅ (added to match the molar ratio in the CYP3A4 + OR + b₅ Supersomes), the enzyme preparation again displayed sigmoidal kinetics with a Hill coefficient of 1.50. The K_m value for the CYP3A4 + OR + b₅ preparation was slightly lower than that of b₅-supplemented CYP3A4 + OR preparation (46.7 versus 67.9 μM), and the V_max of CYP3A4 + OR + b₅ (71.4 nmol/min/nmol) was 3.6-fold higher than that of the b₅-supplemented CYP3A4 + OR (20.1 nmol/min/nmol) (Fig. 1). Due to these significant differences in catalytic activity and because a coexpressed CYP3A5 + OR + b₅ Supersome preparation was commercially unavailable, we chose to compare equivalent CYP3A4 + OR and CYP3A5 + OR preparations supplemented with b₅ (1:3 M ratio) for further kinetic experiments with the different CYP3A probe substrates.

Kinetic Parameters for Substrate Metabolism by CYP3A4 and CYP3A5. Structures of itraconazole and the eight different CYP3A4/5 probe substrates used in this study and sites of metabolism are shown in Fig. 2. The best-fit kinetic model and resulting parameter estimates for the probe substrates are shown in Table 2. For TST biotransformation under the in vitro incubation conditions shown in Table 1, homotropic activation was observed for both CYP3A4 and CYP3A5 (Fig. 3, A and B). The sigmoidal model provided the best fit to the data, yielding the same Hill coefficient of 1.33 for both CYP3A4 and CYP3A5. K_m values for CYP3A4 and CYP3A5 were also similar (66.9 and 51.2 μM, respectively) (Table 2). However, the V_max for CYP3A4 was approximately twice that of CYP3A5 (12.2 versus 5.93 nmol/min/nmol).

ERM N-demethylation catalyzed by CYP3A4 and CYP3A5 was best fit by the simple hyperbolic model. V_max and K_m values for CYP3A4 were 1.45 nmol/min/nmol and 14.7 μM, respectively; for CYP3A5, they were 1.25 nmol/min/nmol and 30.7 μM, respectively (Table 2).

Over the concentration range of 0.5 to 200 μM, MDZ 1′-hydroxylation catalyzed by CYP3A4 and CYP3A5 was also best fit to the simple hyperbolic model (Fig. 3, C and D). The V_max for CYP3A4 was approximately one-fourth that of CYP3A5 (5.18 versus 20.0 nmol/min/nmol), whereas the K_m did not differ greatly (2.82 and 3.56 μM). For the 4-hydroxylation pathway, an Eadie-Hofstee plot of CYP3A5 activity gave a convex curve indicative of substrate inhibition kinetics (not shown), yielding K_si and K_m values of 137 and 11.5 μM, respectively. In contrast, the formation of 4-OH MDZ by CYP3A4 followed the simple hyperbolic model, with a V_max of 1.35 nmol/min/nmol and K_m of 8.44 μM (Table 2).

Homotropic activation of 3-OH FLZ formation was observed for both CYP3A4 and CYP3A5 (data not shown), with Hill coefficients of 1.21 and 2.03. For N-desmethyl FLZ formation, the sigmoidal model provided the best fit for CYP3A5, whereas CYP3A4 followed simple hyperbolic kinetics (Table 2). The K_m value for 3-OH FLZ formation using CYP3A4 was about 2.1-fold higher than that of CYP3A5 (106 versus 49.3 μM); however, for N-desmethyl FLZ, K_m values for CYP3A4 and CYP3A5 were similar (35.6 versus 43.8 μM). V_max parameters for the two enzymes also differed greatly: 3-OH FLZ, 19.3 versus 3.85 nmol/min/nmol; N-desmethyl FLZ, 0.55 versus 0.22 nmol/min/nmol (for CYP3A4 and CYP3A5, respectively).

The metabolism of CBZ by CYP3A4 and CYP3A5 was quite similar. CBZ 10,11-epoxide formation catalyzed by CYP3A4 and CYP3A5 both demonstrated homotropic activation (data not shown), and this was best fit to the sigmoidal model, with large Hill coefficients of 1.75 and 2.09 (high cooperativity), respectively. The K_m values for CYP3A4 and CYP3A5 were 248 and 338 μM, respectively, and the V_max values were 4.87 and 5.98 nmol/min/nmol.

The N-demethylation of DXM by CYP3A4 was best fit by the biphasic saturation model. The Eadie-Hofstee plot showed a concave shape (data not shown) that is usually associated with a two-enzyme kinetic system. The kinetics of 3-MM metabolite formation by CYP3A4 were characterized by high-affinity K_m (56.2 μM), low-capacity V_max (0.84 nmol/min/nmol) and low-affinity K_m (2989 μM), high-capacity V_max (12.1 nmol/min/nmol) processes. In contrast, the formation of 3-MM by CYP3A5 did not seem saturable over the substrate concentration range employed, and only the ratio of V_max/K_m from the linear regression could be estimated.

The formation of the TFN alcohol metabolite by CYP3A4 and CYP3A5 displayed substrate inhibition kinetics (Fig. 3, E and F), yielding K_si values of 287 and 219 μM, respectively. The K_m for CYP3A5 (2.7 μM) was 5.4-fold higher than that of CYP3A4 (0.5 μM). Because of the short incubation period (2 min), no secondary acid metabolite was detected. The V_max values for CYP3A4 and CYP3A5 were 0.86 and 2.06 nmol/min/nmol, respectively (Table 2).

LDC N-demethylation by CYP3A4 and CYP3A5 is shown in Fig. 3, G and H. For both enzymes, the biphasic saturation model provided the best fit to the data. The kinetics of MEGX formation by CYP3A4 was characterized by a high-affinity K_m (78.9 μM), low-capacity V_max (3.99 nmol/min/nmol) and a low-affinity K_m (5927 μM), high-capacity V_max (33.9 nmol/min/nmol). Similar to CYP3A4, LDC N- demethylation catalyzed by CYP3A5 was characterized by high-affinity K_m (21.8 μM), low-capacity V_max (1.51 nmol/min/nmol) and low-affinity K_m (1217 μM), high-capacity V_max (20.6 nmol/min/nmol) processes.

Comparison of CYP3A4 and CYP3A5 Intrinsic Clearances.Table 3 summarizes the calculated in vitro intrinsic metabolite formation clearances for CYP3A4 and CYP3A5. With respect to 6β-OH TST, 3-OH FLZ, N-desmethyl FLZ, and CBZ 10,11-epoxide formation, a maximal intrinsic clearance was calculated using eq. 5 for autoactivation. For 6β-OH TST formation, CYP3A5 was approximately 60% as active as CYP3A4 (0.066 versus 0.104 ml/min/nmol). For 3-OH FLZ and desmethyl FLZ formation, the maximal intrinsic clearances of CYP3A5 were approximately 35 and 20% of those for CYP3A4, respectively. N-Demethylation of ERM also occurred with a lower intrinsic clearance for CYP3A5 (41%) compared with CYP3A4 (0.041 versus 0.099 ml/min/nmol). TFN alcohol formation followed substrate inhibition kinetics. Although the V_max for CYP3A5 was ∼2-fold higher than that of CYP3A4, the maximal intrinsic clearance of CYP3A5 was only 44% of that of CYP3A4. For the remaining substrate/metabolite pairs, CYP3A5 exhibited a maximal intrinsic clearance that was comparable to CYP3A4, including CBZ 10,11-epoxidation, DXM N-demethylation, and LDC N-demethylation (the CYP3A5/CYP3A4 intrinsic clearance ratio was 0.90, 1.07, and 1.35, respectively). For MDZ 1′-hydroxylation, CYP3A5 showed higher activity (3.05-fold) than CYP3A4, whereas CYP3A5 exhibited comparable activity to CYP3A4 (0.88-fold) for the 4-hydroxylation of MDZ.

Metabolite Ratios and Regioselectivity of Midazolam and Flunitrazepam Metabolism. MDZ and FLZ belong to the benzodiazepine class of drugs that undergo CYP3A-mediated hydroxylation at two separate positions: 1′- and 4-OH for MDZ and N-methyl-OH and 3-OH for FLZ. The ratios of major to minor MDZ and FLZ metabolites as a function of substrate concentration are shown in Fig. 4. The ratio of 1′-OH (major) to 4-OH (minor) MDZ formation from CYP3A5 was about 5-fold higher than that for CYP3A4 at the lowest MDZ concentration (0.5 μM). For CYP3A5, the ratio was 35.9 at 0.5 μM and decreased to a minimum of 16.9 at 20 μM. However, due to apparent substrate inhibition of 4-OH MDZ formation, the ratio increased to 28.5 at 200 μM. In contrast, the ratio of 3-OH (major) to N-desmethyl (minor) FLZ for CYP3A5 was similar to the corresponding ratio for CYP3A4 at the low concentration of 20 μM (10.8 versus 11.1). For CYP3A4, this ratio increased dramatically to a maximum of 31.6 at 500 μM, whereas for CYP3A5 the ratio increased to a maximum of 18.4 at 300 μM.

Expression Levels of CYP3A, Reductase, and b₅ in Selected HLMs.Table 4 summarizes the expression levels of CYP3A4, CYP3A5, P450 reductase, and cytochrome b₅ in group 1 and 2 human liver microsomes. Microsomes for group 1 and 2 were selected to differ significantly in CYP3A5 content (p < 0.001), whereas their mean CYP3A4 contents were similarly matched. Moreover, each of the 10 individual pairs from group 1 and 2 differed in CYP3A4 content by less than 10% except for one pair that had a 25% difference. Accordingly, the mean total CYP3A content of group 2 was 2.2-fold higher than that of group 1. There was no significant difference in the mean expression levels of reductase and b₅ between the two groups (p > 0.05). Although CYP3A contents in the 10 paired HLMs ranged from 54.5 to 381 pmol/mg (7-fold), reductase and b₅ coenzyme contents were more restricted, varying from 36 to 56 pmol/mg (1.6-fold) and 0.26 to 0.52 nmol/mg (2-fold), respectively. Thus, although we could not match reductase and b₅ levels for each of the 10 paired samples, the extent of variability was limited.

Substrate Metabolism by Matched Human Liver Microsomes.Table 5 summarizes the formation rates for 11 metabolites of the nine different CYP3A4/5 probes. Metabolic activities were determined at a substrate concentration ≤K_m to permit approximation of an in vitro intrinsic clearance in the microsomal system.

For all probe substrates, the mean catalytic activity of group 2 microsomes containing relatively high levels of CYP3A5 was greater than that of group 1 microsomes. The ratio of group 2/group 1 metabolic activity was 2.2 for FLZ oxidation (p < 0.001); 1.9 for total MDZ oxidation (p < 0.05); 1.9 for ERM N-demethylation (p < 0.05); 1.8 for DMX N-demethylation (p < 0.05); 1.7 for TST 6β-hydroxylation (p < 0.05); 1.7 for CBZ 10,11-epoxidation (not significant); 1.4 for TFN C-oxidation (not significant); and 1.5 for LDC N-demethylation (not significant).

Results from previous experiments showed that ITZ is an excellent CYP3A4 substrate but is not metabolized by CYP3A5 (Isoherranen et al., 2004). For this study, we examined its metabolism by group 1 and 2 liver microsomes under saturating ITZ concentrations. When probe catalytic activity was adjusted for ITZ activity (for each liver, the nominal probe reaction rate/OH-ITZ formation rate), the results showed that the product formation activity was significantly (p < 0.05) higher in group 2 than in group 1 for the three substrates, and the metabolite formation rate ratios of group 2/group 1 were 1.35, 1.36, and 1.54 for ERM N-demethylation, MDZ oxidation, and FLZ oxidation, respectively (Fig. 5). CYP3A5 still accounted for 26, 27, and 35% of ERM N-demethylation, MDZ oxidation, and FLZ oxidation, respectively.

Table 6 summarizes bivariate Pearson correlations of the various CYP3A substrates for the combined 20 livers tested, irrespective of CYP3A5 status. TST, ERM, MDZ, and ITZ were most highly correlated with each other (r ≥ 0.87). In addition, TST, ERM, MDZ, ITZ, FLZ, CBZ, and DXM metabolite formation rates were all well correlated with each other (r ≥ 0.78). The correlation between the rate of TFN and LDC metabolism and that of the other CYP3A substrates was much weaker (r = 0.40-0.66). All correlations except that between TFN and LDC were significant.

Discussion

An intronic mutation in the CYP3A5 gene (A6986G; CYP3A5^*3) explains in large part the polymorphic pattern of enzyme expression found in human liver, small intestine, and kidney (Hustert et al., 2001; Kuehl et al., 2001; Lin et al., 2002; Givens et al., 2003). However, there is considerable controversy as to whether the expression of CYP3A5 in people carrying the wild-type CYP3A5^*1 gene contributes in a clinically meaningful way to the metabolic clearance of CYP3A substrates. Results from the current study suggest that, for some substrates, the contribution could be significant. Recombinant CYP3A5 enzymes supplemented with cytochrome b₅ exhibited higher or comparable intrinsic metabolite formation clearances for midazolam 1′-hydroxylation, lidocaine N-demethylation, carbamazepine 10,11-epoxidation, and dextromethorphan N-demethylation compared with CYP3A4. For the first three drugs (MDZ, LDC, and CBZ), the metabolic pathway studied represents the dominant route of elimination in vivo. In addition, CYP3A5 catalyzed the major route of terfenadine (C-oxidation) and erythromycin (N-demethylation) elimination with intrinsic clearances that were approximately 40% that of CYP3A4. CYP3A5 was also found to be a good catalyst of testosterone 6β-hydroxylation; the intrinsic formation clearance was 60% that of CYP3A4. Even the worst substrate probe for heterologously expressed CYP3A5, flunitrazepam, was metabolized with an efficiency that was 20 to 35% that of CYP3A4.

Our experimental findings are quite different from those reported recently by two independent research groups. Williams et al. (2002) found that, except for midazolam, CYP3A5 was a much poorer catalyst of drug metabolism, including substrates such as testosterone (CYP3A5/CYP3A4 CL_int ratio = 0.03), which was included in our data set. However, those investigators used a CYP3A4 preparation that was coexpressed with reductase and b₅ and a CYP3A5 preparation that required supplementation of both b₅ and reductase. Patki et al. (2003) also compared the catalytic activity of CYP3A4 and CYP3A5 heterologously expressed in baculovirus-transfected insect cells with a comparable level of reductase activity but no b₅ (overexpressed or supplemental) and again reported much lower metabolic activity for the CYP3A5 enzyme compared with CYP3A4. A major difference between these two studies and our own pertains to the enzyme preparations employed and how they were supplemented with coenzymes. We adopted our experimental approach based on the work of Hirota et al. (2001), who also compared alprazolam α-hydroxylation activity from CYP3A4 that was either coexpressed or supplemented with b₅. Their results revealed a 23-fold higher intrinsic clearance for the CYP3A4 preparation that was coexpressed with b₅ compared with CYP3A4 that was supplemented with b₅. We found a similar, although lower, magnitude difference in our pilot study. Testosterone 6β-hydroxylation activity from coexpressed CYP3A4/reductase/b₅ was more than 3-fold higher than CYP3A4/reductase that required supplementation with extra reductase and b₅ to achieve the same molar ratios. Because CYP3A5 coexpressed with P450 reductase and cytochrome b₅ was not available for comparison with the “optimal” CYP3A4 enzyme preparation, we chose a different strategy.

There are additional published data from incubations with heterologously expressed enzymes that demonstrate the comparable metabolic activity for CYP3A4 and CYP3A5. For example, Gillam et al. (1995) compared recombinant CYP3A4 and CYP3A5 that had been purified and reconstituted in an identical manner with reductase, cytochrome b₅, and lipid matrix adjuvants that optimized catalytic activity. They found an excellent (maximal) activity profile for CYP3A5, compared with CYP3A4, for the substrates nifedipine, testosterone, N-ethylmorphine, erythromycin, and d-benzphetamine. Moreover, using commercial preparations of recombinant enzyme, Lee et al. (2001) found that CYP3A5, compared with CYP3A4, exhibited a similar rate of 4-hydroxy-17β-estradiol formation and a rate of 2-hydroxy-17β-estradiol formation that was approximately one-third that of CYP3A4. CYP3A5 was also an excellent catalyst of fentanyl oxidation but less so for alfentanil and sufentanil compared with CYP3A4 (Guitton et al., 1997). Compared with CYP3A4, recombinant CYP3A5 was also a good catalyst of diltiazem N-demethylation (Yamaori et al., 2004) and estrone metabolism, although again there were interesting differences in product regioselectivity for the two enzymes (Lee et al., 2002).

The relative metabolic activity of any recombinant enzyme may not accurately reproduce its catalytic function in vivo. Thus, a major and novel strength of this paper is the supporting results obtained from CYP3A protein phenotyped liver microsomes. Human livers from a tissue bank were preselected based on immunodetectable CYP3A4 and CYP3A5 contents to achieve two groups of 10 livers each that contained comparable CYP3A4 levels but markedly different CYP3A5 content; group 1 had a low CYP3A5 content, and group 2 had a high CYP3A5 content. Using substrate concentrations below or equal to the K_m revealed that five of the probe molecules were metabolically cleared more rapidly by those livers with readily detectable levels of CYP3A5 (CYP3A5^*1 heterozygous or homozygous genotype), compared with those with a homozygous mutant CYP3A5^*3 genotype. However, when we used ITZ hydroxylation activity to control for possible interliver variability in specific CYP3A4 activity (activity/nanomoles of detectable protein), we found a significant difference between group 2 and 1 microsomal activity toward only ERM, MDZ, and FLZ (p < 0.05). The adjusted results showed that, for these drugs, CYP3A5 on average accounted for 26, 27, and 35% of the total product formation catalyzed by group 2 (CYP3A5^*1) microsomes, respectively.

The significant results obtained for liver microsomal FLZ and ERM metabolism are somewhat surprising, given that the intrinsic clearance of recombinant CYP3A5 was only 20 to 35% and 41% that of recombinant CYP3A4. We have no clear explanation for the discrepancies other than to attribute them to limitations of the in vitro experimental systems or possibly to uncontrolled variability in other metabolizing enzymes such as CYP2C19 for FLZ (Kilicarslan et al., 2001). Our experimental findings combined with some, but not all, recently published results suggest that the CYP3A5 polymorphism can contribute importantly to interindividual variability in metabolic drug clearance in vivo, but its impact will be highly substrate-dependent. This interpretation is supported by investigations of tacrolimus disposition in organ transplant patients. Results from every study reported to date reveal higher steady-state blood tacrolimus concentration/dose ratios (lower oral clearance) in subjects with a homozygous CYP3A5^*3 genotype compared with those carrying the functional CYP3A5^*1 gene (Hesselink et al., 2003; Thervet et al., 2003; Zheng et al., 2003). Surprisingly, the published data for midazolam are more equivocal. Results reported by some investigators suggest higher metabolic clearances for CYP3A5^*1 subjects (Goh et al., 2002; Wong et al., 2004), whereas other investigators have failed to find such an association (Shih and Huang, 2002; Floyd et al., 2003; Yu et al., 2004).

A mechanistic explanation for the in vitro/in vivo discrepancy for midazolam clearance is not readily apparent, given that the drug is an excellent substrate for CYP3A5. It is possible that unrecognized mutations in the CYP3A genes have confounded interpretation of in vivo study results. In addition, the in vivo studies conducted to date may be confounded by uncontrolled interindividual variability in CYP3A4 activity and an insufficient study sample size. It is also possible that non-CYP3A factors (such as plasma protein binding or hepatic/intestinal blood flow) contribute significantly to the variability in midazolam clearance and bioavailability in addition to tissue CYP3A4/5 content.

Katz et al. (2004) have offered an additional explanation for the conflicting in vivo penetrance of the CYP3A5 mutations. In their study of the kinetics of ABT-773 in healthy volunteers, they found that the impact of the polymorphism on the area under the curve of ABT-773 was dose-dependent. Subjects carrying the CYP3A5^*1 allele showed a significantly lower area under the curve than did subjects with two inactivating alleles (CYP3A5^*3 or CYP3A5^*6), but only at the highest (450 mg) oral dose employed. The authors suggested that this observation may reflect saturable ABT-773 kinetics and the lower-affinity and high-capacity nature of CYP3A5 enzyme kinetics compared with CYP3A4 (a characteristic observed for some of our substrates). They concluded that the impact of the CYP3A5 polymorphism for any CYP3A substrate may become more apparent when it is dosed to achieve CYP3A4 saturating concentrations in the small intestine and liver. Considering the K_m and V_max values we report in this study, oral erythromycin and carbamazepine might be suitable for further evaluation of a dose-dependent pharmacogenetic effect in vivo. The hypothesis posited by Katz et al. (2004) is quite intriguing, particularly when considering drugs with a high intestinal first-pass effect. However, it does not help explain the consistent effect of the CYP3A5 mutations on oral tacrolimus disposition, given that the K_m and V_max for tacrolimus disappearance in incubations with CYP3A4 or CYP3A5 are similar (Y. Dai, K. Iwanaga, Y.S. Lin, M.F. Hebert, C.L. Davis, and K.E. Thummel, unpublished results).

There were other aspects of CYP3A4 and CYP3A5 kinetics that show a high degree of commonality. Both CYP3A4 and CYP3A5 displayed homotropic activation kinetics for the 6β-hydroxylation of testosterone, 3-hydroxylation of flunitrazepam, and 10,11-epoxidation of carbamazepine; biphasic saturation for the N-demethylation of lidocaine; substrate inhibition for terfenadine C-hydroxylation; and simple hyperbolic saturation for midazolam 1′-hydroxylation and erythromycin N-demethylation. Evidence for CYP3A5 heteroactivation has been reported by other investigators (Ueng et al., 1997); thus, our finding of autoactivation for testosterone, flunitrazepam, and carbamazepine is not unexpected. Although the product formation kinetics for midazolam 1′-hydroxylation that were observed in this study (simple hyperbolic) differ from the substrate inhibition kinetics sometimes described in the published literature (Kronbach et al., 1989), this difference may reflect the restricted incubation conditions (time and concentration) we employed to limit the possibility of suicide inhibition by midazolam (Schrag and Wienkers, 2001; Khan et al., 2002a) or a difference in enzyme source.

The allosteric interactions of CYP3A4 and CYP3A5 substrates were not completely concordant. Both midazolam and flunitrazepam exhibited a concentration-dependent change in the ratio of the major and minor oxidation products. For midazolam, CYP3A5 gave a higher 1′-OH/4-OH product ratio at low substrate concentrations and a more pronounced decrease in the ratio with increasing concentrations than did CYP3A4. However, for both enzymes, metabolic switching to the minor metabolite occurred. Conversely, for flunitrazepam, the baseline 3-OH/N-desmethyl ratio of CYP3A5 was similar to that of CYP3A4, but it increased to a lesser degree than the product ratio from CYP3A4. For both enzymes, dominance of the 3-OH metabolite was enhanced rather than decreased, and the effect was more pronounced for CYP3A4. Based on current theory, these differences in product ratio reflect different binding affinities for the multiple substrate molecules that fit into a CYP3A active site (Korzekwa et al., 1998) or binding at different sites of the enzyme (Ueng et al., 1997), possibly as a result of key amino acid differences at substrate and effector sites (Khan et al., 2002a,b).

In summary, the data we report suggest that CYP3A5 can make a substantial contribution to the total metabolic intrinsic clearance of some drugs. The extent of contribution will be highly dependent on the presence of at least one copy of the wild-type CYP3A5^*1 allele. It will also depend on the substrate under consideration. The balance of our in vitro evidence indicates that midazolam and erythromycin and possibly flunitrazepam should be affected by the polymorphic CYP3A5 expression in vivo. A further in vivo pharmacogenetic investigation of these substrates and others is warranted, using (where possible) a study design that probes both the dose and route of administration effects.